Underwater Image Enhancement Using Color Constancy Via Homomorphic Filtering and Depth Estimation

Underwater image visibility depends on the light scattering and absorption, and the illumination variations degrade the image details with color distortion, low contrast, blurring, non-uniform illumination etc. An image enhancement framework is proposed to address the degradation problems in underwater images. This framework comprises three steps: The first step is to process an input RGB image via white patch retinex (WPR) based gamma correction for illumination compensation and chromatic adaptation technique (CAT) in adjusting the color in RGB image based on estimated scene illuminant by WPR; Second step is to perform image enhancement via homomorphic filtering by considering only the high-frequency components with high frequency emphasis filter (HFEF), and finally dark channel prior (DCP) method is adapted to obtain the visually enhanced image by preserving the image details. The proposed method is tested on a standard underwater image dataset, and observed that qualitative and quantitative evaluations are better with visibility and color restoration with reduced artifacts and halo effects.